A method for human-machine collaborative disassembly of retired power battery based on digital twin assistance

Abstract

The application discloses a kind of based on digital twin auxiliary's retired power battery man-machine collaborative disassembly method, comprising: obtaining the 3D feature data of physical model, 3D feature data is transmitted to digital twin platform;According to 3D feature data, construct digital model, and realize real-time monitoring to physical model by real-time data intermingling;According to the CAD model of retired power battery, call disassembly analysis algorithm, obtain the basic disassembly information of the battery, assess the disassembly of the battery, determine disassembly sequence;According to disassembly sequence, and the disassembly capacity of operator and robot, construct dynamic bayesian network, dynamically allocate disassembly task, obtain the best task allocation;According to the best task allocation, carry out disassembly simulation in digital twin platform, determine the feasibility of the scheme;According to disassembly simulation result, planning robot motion path, complete man-machine collaborative disassembly.The application realizes man-machine collaborative disassembly, improves the efficiency of disassembly process.

Description

Technical Field

[0001] This invention relates to the field of power battery dismantling, and in particular to a human-machine collaborative dismantling method for retired power batteries based on digital twin assistance. Background Technology

[0002] Today, many remanufacturing industries have a huge demand for recycling processes to improve the utilization of valuable resources. Disassembly is a crucial step in the recycling process. In recent years, robotics has been increasingly applied to disassembly processes to complete many complex remanufacturing tasks. However, due to the high complexity and diversity of retired products, manual disassembly is not economically efficient, and robotic systems exhibit high uncertainty when handling complex disassembly operations, leading to unreliable disassembly.

[0003] During disassembly, both humans and robots can handle tasks independently and interact with each other in various ways. Therefore, there is growing interest in human-robot collaboration. In recent years, Human-Robot Collaboration (HRC) has opened new doors for remanufacturing and disassembly processes due to its advantages in flexibility and efficiency. HRC allows humans and robots to work closely together on the same task. Human operators can flexibly utilize cognitive knowledge that robots lack to solve problems, while robots, as collaborating partners, can leverage their strengths in power, precision, and repeatability to assist humans. HRC improves system efficiency and enhances overall productivity.

[0004] However, unlike the assembly process, disassembly is influenced by various factors and faces greater uncertainty, including the product's disassembly feasibility and the difficulty of operation. When robots and humans collaborate, the uncertainty and complexity of the disassembly process increase dramatically.

[0005] In recent years, digital twin technology, as an emerging intelligent technology, has enabled the docking and interaction between entities in the physical world (such as devices, components, processes, and people) and their corresponding models in the virtual world (such as modeling, simulation, and data analysis). Through real-time data acquisition, model generation, and analysis, the behavior and state of objects in the physical world can be reflected in real time, providing proactive support for operational decisions, thereby enabling the connection, simulation, and control of physical entities. Currently, digital twin technology is widely used in human-machine collaboration, aiming to support the design, construction, and control of human-machine collaboration. Real-time monitoring of digital twin platforms provides a new approach to solving uncertainties in the disassembly process of human-machine collaboration. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a human-machine collaborative dismantling method for retired power batteries based on digital twin assistance, which addresses the deficiencies in the existing technology.

[0007] The technical solution adopted by this invention to solve its technical problem is:

[0008] This invention provides a human-machine collaborative dismantling method for retired power batteries based on digital twin assistance, the method comprising the following steps:

[0009] S101. Acquire 3D feature data of the physical model of the retired power battery and robot by image capture, and transmit the 3D feature data to the digital twin platform.

[0010] S102. Construct a digital model based on 3D feature data, and achieve real-time monitoring of the physical model through real-time data fusion;

[0011] S103. Based on the CAD model of the retired power battery, call the disassembly analysis algorithm to obtain the basic disassembly information of the battery, evaluate the disassembly feasibility of the battery, and determine the disassembly sequence.

[0012] S104. Based on the disassembly sequence and the disassembly capabilities of the operator and the robot, construct a dynamic Bayesian network to dynamically allocate disassembly tasks and obtain the optimal task allocation.

[0013] S105. Based on the optimal task allocation, conduct a breakdown simulation on the digital twin platform to determine the feasibility of the solution;

[0014] S106. Based on the disassembly simulation results, plan the robot's motion path and complete the human-machine collaborative disassembly.

[0015] Further, the method of step S101 of the present invention includes:

[0016] In the real world, 3D feature data of physical models of retired power batteries and robots are collected by 3D motion-sensing cameras and transmitted to a digital twin platform based on the Unity game engine.

[0017] In a digital twin platform, virtual models and virtual scenes are constructed based on the 3D feature data of the physical model. Unity's scripting capabilities are used to write code to achieve data interaction, data analysis, and real-time data monitoring.

[0018] Further, the method of step S103 of the present invention includes:

[0019] The digital twin platform will utilize disassembly analysis algorithms to analyze the CAD model of the retired power battery and obtain its disassembly tree. Based on this tree, a preliminary disassembly sequence will be determined. Further analysis of each disassembly node will then be conducted, and its disassembly capability will be comprehensively evaluated based on its connection method and disassembly capacity. Following this, the overall disassembly capability of the battery will be comprehensively evaluated based on the preliminary disassembly sequence and the analysis results of the disassembly capability of each node. Finally, based on the battery's disassembly capability, the disassembly sequence for removable batteries will be determined, and non-removable batteries will be either crushed as a whole or reused in a tiered manner.

[0020] Furthermore, the method of step S103 of the present invention specifically includes:

[0021] (1) Analyze the CAD model of the retired power battery. Based on the three-dimensional spatial characteristics of the retired power battery in the CAD drawings and the disassembly experience of the retired power battery, obtain the disassembly tree of the retired power battery. Then, based on the disassembly tree of the retired power battery, preliminarily determine the disassembly sequence of the battery:

[0022]

[0023] (2) Further analysis of each dismantling node, and analysis of the connection method of the retired power battery node:

[0024]

[0025] In the formula, c ii N represents the elements on the main diagonal of matrix C. All other elements in matrix C are assumed to be 0. j This indicates the number of nodes; the methods for determining the difficulty of node decomposition include:

[0026] When the connection method of the retired power battery node is bolt connection, spring connection, or mechanical pressing, the node is considered easy to disassemble; when the connection method is welding connection, the node is considered difficult to disassemble; when the key nodes related to the battery unit are difficult to disassemble, the disassembly and recycling of the retired power battery is not very meaningful, and matrix C is assigned the value of zero matrix.

[0027] (3) Analyze the real-time status of retired power battery nodes:

[0028]

[0029] In the formula, d ii N represents the elements on the main diagonal of matrix C. Other elements in matrix D are assumed to be 0. j Indicates the number of nodes; when the node's shape... When the deformation exceeds the deformation threshold by 10%, the node is considered to have a large degree of rusting and deformation. When the deformation Δ of the key node related to the battery cell exceeds the deformation threshold, it is not very meaningful for the dismantling and recycling of the retired power battery. At this time, matrix D is assigned the value of zero matrix.

[0030] (4) Taking into account the retired power battery node situation reflected by matrices C and D, when r(C) + r(D) ≥ 2N j When the value is -3 and r(C)≠0, r(D)≠0, the retired power battery is considered to be disassembled; where r(C) and r(D) are the ranks of matrix C and matrix D, respectively.

[0031] (5) Based on the disassembly of the retired power battery, the initially determined disassembly sequence is adjusted to obtain the final disassembly sequence.

[0032] Further, the method of step S104 of the present invention includes:

[0033] Based on the disassembly sequence and the disassembly capabilities of the operator and the robot, a dynamic Bayesian network is constructed. According to the characteristics of discrete Bayesian networks, the dynamic Bayesian network is converted into a hidden Markov model. The hidden Markov model is then improved according to the actual situation, and the disassembly tasks are dynamically allocated.

[0034] An improved Hidden Markov Model (HMM) is constructed. In the task allocation for dismantling retired power batteries, the operator working at time t is defined as the hidden sequence Q of the improved HMM, and the operation in progress at time t is defined as the observation sequence V of the improved HMM.

[0035]

[0036] V = {v1, v2, ... v} M}

[0037] The key parameters of the improved Hidden Markov Model are determined, λ = (π, A, B); π represents the initial state probability vector, which is used in the task allocation for dismantling retired power batteries. i This represents the initial probability of selecting either a human operator or a robot to perform the disassembly task;

[0038] π=(π i ),π i =P(i i =q1), i = 1, 2, ..., N o

[0039] A is from state q i Transition to state q j The state transition probability matrix, in the task allocation for dismantling retired power batteries, represents the state transition probability matrix at time t for operator a. ijAt time t+1, B represents the probability of a certain operator; B is derived from state q. j Observing a specific state v k The probability distribution matrix, in the allocation of tasks for dismantling retired power batteries, represents the probability distribution matrix when the operator is b. ij At that time, b ij The ongoing dismantling task represents the probability of a certain dismantling task.

[0040] An improved Hidden Markov Model is employed, combined with expert knowledge or experience, and parameter estimation based on Bayes' theorem is used to determine the state transition probability matrix and probability distribution matrix:

[0041] (1) Define the prior probability density function

[0042] Based on expert knowledge or experience, define the prior probability density functions of the state transition probability matrix and the probability distribution matrix:

[0043] p(A|α d )

[0044] p(B|β d )

[0045] Where α d β d It is a probability distribution for the dismantling of retired power batteries derived from expert knowledge or experience;

[0046] (2) Calculate the posterior probability density function

[0047]

[0048] Where p(V|A,B) represents the probability of the observed sequence V given model parameters A and B, i.e., during the dismantling of retired power batteries, given time t, the operator a... ij At time t+1, the state probability transition matrix A is for a certain operator, and the matrix A is for when the operator is b. ij At that time, b ij When the ongoing dismantling task is the probability distribution matrix B of a certain dismantling task, the observation sequence V is the probability of the dismantling operation in progress at time t.

[0049] Normalize the posterior probability density function:

[0050]

[0051] Among them, A i B j Representing the possible values ​​of A and B, after normalization, we get This represents the probability that A and B take different values ​​when given an observation sequence V during the dismantling of retired power batteries;

[0052] (3) Determine the state transition probability matrix A and the probability distribution matrix B based on the posterior probability density function:

[0053]

[0054] Where, α t (i,j) represents the posterior probability that the user is in state i at time t and in state j at time t+1, where N is the posterior probability. o This indicates the number of elements in the hidden sequence, i.e., the number of operators involved in the dismantling of retired power batteries;

[0055]

[0056] Among them, v k β represents the k-th observation in the observation sequence. t (i) represents the posterior probability of being in state i at time t, N o M represents the number of elements in the hidden sequence, i.e., the number of operators in the dismantling process of retired power batteries; M represents the number of elements in the observed sequence, i.e., the number of dismantling tasks in the dismantling process of retired power batteries.

[0057] Next, the task allocation is used as the hidden sequence Q, and the output decomposition sequence is used as the observation sequence V. The Viterbi algorithm is then used to calculate the optimal task allocation.

[0058] Furthermore, the specific method of the present invention for calculating the optimal task allocation using the Viterbi algorithm includes:

[0059] The Hidden Markov Model (HMM) is input with parameters λ = (π, A, B) and the observation sequence V. The Viterbi algorithm is then used to decode and analyze the model, ultimately yielding the optimal task allocation. As an improvement, constraint information between states is introduced during the Viterbi algorithm's recursion process to eliminate some unreasonable task allocations and simplify computation. This includes:

[0060] (1) Initialization:

[0061] δ1(i)=π i b i (v1), i = 1, 2, ..., N o

[0062] ψ1(i)=0,i=1,2,…,N o

[0063] (2) Recursively calculate t = 2, 3, ..., T:

[0064] Define constraint matrix P

[0065]

[0066] Where, N m The number of decomposition operations is indicated by the constraint matrix P, which is then introduced into the recursive process.

[0067]

[0068]

[0069] (3) Termination:

[0070]

[0071]

[0072] Where, δ t (i) represents the maximum probability of all individual paths to state i at time t, i.e., the operator at time t in the dismantling process of the retired power battery is a. ij At that time, the maximum probability of all individual tasks being assigned a path; ψ t (i) represents the (t-1)th node of the path with the highest probability among all individual paths at time t, i.e., the operator in the dismantling process of the retired power battery at time t is a. ij At time t-1, the operator a of the path with the highest probability among all individual task assignment paths. i′j′ .

[0073] Further, the method of step S105 of the present invention includes:

[0074] Based on the calculated maximum δ at time T T (i) is the probability of the most likely hidden state occurring, i.e., the probability of optimal task allocation during battery disassembly; calculate the maximum ψ at time T. T (i) is the most likely hidden state at time T, i.e., the optimal task allocation at time T; finally, ψ is used. T (i) Backtracking yields the most likely hidden sequence Q, which is the optimal task assignment:

[0075] Q = {Robot 1, Operator 1, Robot 1, Robot 2, ...}.

[0076] This invention provides a human-machine collaborative dismantling system for retired power batteries based on digital twin assistance, comprising:

[0077] The physical model in the real environment is used to acquire 3D feature data of the physical model of the retired power battery and robot through image capture, and the 3D feature data is transmitted to the digital twin platform.

[0078] The digital twin platform includes:

[0079] Virtual models are used to interact with physical models in real time, enabling real-time monitoring of the physical models through real-time data fusion.

[0080] The disassembly analysis algorithm is used to obtain basic disassembly information of the battery based on the CAD model of the retired power battery, evaluate the battery's disassembly feasibility, and determine the disassembly sequence. Based on the disassembly sequence and the disassembly capabilities of the operator and robot, a dynamic Bayesian network is constructed to dynamically allocate disassembly tasks and obtain the optimal task allocation.

[0081] Disassembly simulation: Based on the optimal task allocation, disassembly simulation is conducted on the digital twin platform to determine the feasibility of the solution; based on the disassembly simulation results, the robot's motion path is planned to complete the human-machine collaborative disassembly.

[0082] Furthermore, the digital twin platform of the present invention includes a disassembly preprocessing module, which is used to call a disassembly analysis algorithm to analyze the CAD model of the retired power battery and obtain the disassembly tree of the battery; then, based on the disassembly tree of the battery, the disassembly sequence of the battery is initially determined; then, each disassembly node is further analyzed, and the disassembly feasibility of each node is comprehensively evaluated based on the connection method and disassembly capability of each node; then, based on the initially determined disassembly sequence and the disassembly feasibility analysis results of each node, the overall disassembly feasibility of the battery is comprehensively evaluated; finally, based on the disassembly feasibility of the battery, the disassembly sequence of the disassembly of the disassemblyable batteries is determined, and the non-disassemblyable batteries are subjected to whole-machine crushing or cascade utilization.

[0083] Furthermore, the digital twin platform of the present invention includes a task planning module, which is used to construct a dynamic Bayesian network based on the disassembly sequence and the disassembly capabilities of the operator and the robot, convert the dynamic Bayesian network into a hidden Markov model based on the characteristics of the discrete Bayesian network, and improve the hidden Markov model according to the actual situation, and dynamically allocate disassembly tasks.

[0084] The digital twin platform includes a disassembly simulation module, which is used to perform disassembly simulation based on the disassembly task allocation calculated by a dynamic algorithm. The module assigns specific disassembly tasks to operators and robots in the virtual environment, simulates the actual disassembly process, eliminates potential safety hazards, and determines the feasibility of the solution.

[0085] The beneficial effects of this invention are:

[0086] 1. Compared with existing technologies, this invention uses a digital twin platform to assist human-machine collaborative disassembly, leveraging the advantages of digital twin technology in real-time simulation monitoring. At the same time, the digital twin platform integrates a disassembly analysis module to analyze the product's disassembly feasibility, disassembly strategy, and task allocation in a virtual scenario, and performs disassembly simulation to obtain the optimal human-machine collaborative disassembly task allocation, thus solving problems such as real-time monitoring of the disassembly process and dynamic allocation of disassembly tasks.

[0087] 2. This invention combines the operator's flexibility with the robot's reliability in handling problems through a digital twin-assisted platform, giving full play to the advantages of robotics technology and realizing human-machine collaborative disassembly, providing a new approach to improving the efficiency of the disassembly process.

[0088] 3. A disassembly analysis algorithm is proposed to analyze the CAD model of the retired power battery and obtain the battery disassembly tree. Then, based on the battery disassembly tree, the disassembly sequence of the battery is initially determined. Then, each disassembly node is further analyzed, and the disassembly feasibility of each node is comprehensively evaluated. This algorithm can effectively and accurately evaluate the disassembly feasibility of the battery.

[0089] 4. In the dismantling task planning algorithm, the classic Hidden Markov Model usually uses maximum likelihood estimation to determine the model parameters. However, in the Hidden Markov Model for power battery dismantling, maximum likelihood estimation is easily affected by data overfitting. Therefore, this invention adopts an improved Hidden Markov Model, combined with expert knowledge or experience, and uses parameter estimation based on Bayes' formula to determine the state transition probability matrix and probability distribution matrix.

[0090] 5. When calculating the optimal task allocation, the Viterbi algorithm was improved by introducing constraint information between states during the recursive process of the Viterbi algorithm, eliminating some unreasonable task allocations and simplifying the computation.

[0091] 6. During the disassembly simulation, based on the disassembly task allocation calculated by the dynamic algorithm, the disassembly simulation is carried out on the digital twin platform. The specific disassembly tasks are assigned to the operators and robots in the virtual environment to simulate the actual disassembly process, eliminate potential safety hazards, and determine the feasibility of the solution. Attached Figure Description

[0092] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings:

[0093] Figure 1 This is a flowchart of an embodiment of the present invention;

[0094] Figure 2 This is a schematic diagram of the structure of the digital twin platform according to an embodiment of the present invention;

[0095] Figure 3This is a flowchart of a human-machine collaborative disassembly and analysis according to an embodiment of the present invention;

[0096] Figure 4 This is a flowchart of the disassembly process according to the disassembly sequence and the disassembly process of the operator and the robot in an embodiment of the present invention;

[0097] Figure 5 This is a schematic diagram of the disassembly task allocation in an embodiment of the present invention. Detailed Implementation

[0098] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0099] Example 1

[0100] like Figure 1 As shown in the figure, this embodiment of the invention provides a human-machine collaborative dismantling method for retired power batteries based on digital twin assistance. The method includes the following steps:

[0101] S101. Acquire 3D feature data of the physical model of the retired power battery and robot by image capture, and transmit the 3D feature data to the digital twin platform.

[0102] S102. Construct a digital model based on 3D feature data, and achieve real-time monitoring of the physical model through real-time data fusion;

[0103] S103. Based on the CAD model of the retired power battery, call the disassembly analysis algorithm to obtain the basic disassembly information of the battery, evaluate the disassembly feasibility of the battery, and determine the disassembly sequence.

[0104] S104. Based on the disassembly sequence and the disassembly capabilities of the operator and the robot, construct a dynamic Bayesian network to dynamically allocate disassembly tasks and obtain the optimal task allocation.

[0105] S105. Based on the optimal task allocation, conduct a breakdown simulation on the digital twin platform to determine the feasibility of the solution;

[0106] S106. Based on the disassembly simulation results, plan the robot's motion path and complete the human-machine collaborative disassembly.

[0107] In a preferred embodiment of the present invention, the present invention utilizes digital twin technology to assist human-machine collaborative disassembly, giving full play to the advantages of digital twin technology in real-time monitoring.

[0108] like Figure 2 As shown, this invention provides a digital twin-assisted platform. Figure 2Only some of the main components are shown. In the figure, 200 represents the intelligent human-machine collaborative dismantling system for retired power batteries assisted by digital twins, 201 represents the real environment, and 202 represents the digital twin platform.

[0109] Combination Figure 2 This invention utilizes the Microsoft Kinect V2 3D motion-sensing camera to collect 3D feature parameters of a model in the physical world, and transmits the 3D feature data to a Unity-based digital twin platform via sensors. Then, based on the 3D features of the physical model, a virtual model and virtual scene are constructed. Unity's scripting capabilities are used to write code for data interaction, data analysis, and real-time data monitoring. Next, through the Unity-based digital twin platform, a disassembly analysis algorithm is invoked to analyze the CAD model of a retired power battery and the characteristics of each node, obtaining basic disassembly information, assessing the battery's disassembly feasibility, and determining the disassembly sequence. Then, based on the disassembly sequence and the disassembly capabilities of humans and robots, a dynamic Bayesian network is constructed, and the Viterbi algorithm is used to dynamically allocate disassembly tasks to obtain the optimal task allocation. Furthermore, based on the disassembly task allocation calculated by the dynamic algorithm, a disassembly simulation is performed on the Unity-based digital twin platform to eliminate safety hazards and determine the feasibility of the solution. Finally, based on the disassembly simulation results, the task allocation instructions are sent to the real environment, the robot's movement path is planned, and human-robot collaborative disassembly is completed.

[0110] In this invention, the advantages of digital twins mainly lie in real-time simulation monitoring and dynamic task allocation. A disassembly analysis module is integrated into the Unity-based digital twin platform. This module analyzes the product's disassembly feasibility, disassembly strategies, and task allocation in a virtual scene, and performs disassembly simulations to derive the optimal human-machine collaborative disassembly task allocation. The control platform sends task allocation instructions to the real-world environment, controlling the robot's movement trajectory to coordinate with the operator, thus achieving human-machine collaborative disassembly.

[0111] Example 2

[0112] like Figure 3 The diagram shown is a flowchart of a human-machine collaborative disassembly and analysis process, illustrating only the main components. In the diagram, 300 represents the main modules of the digital twin platform, 301 is the disassembly preprocessing module, 302 is the disassembly task planning module, and 303 is the disassembly simulation module.

[0113] Based on the CAD model of the retired power battery, a disassembly analysis algorithm is invoked to obtain basic disassembly information of the battery, assess its disassembly feasibility, and determine the disassembly sequence, including:

[0114] Disassembly Pre-processing Module: For retired power batteries, the digital twin platform will call disassembly analysis algorithms to analyze the CAD model of the retired power battery and obtain the disassembly tree of the battery. Then, based on the disassembly tree of the battery, the disassembly sequence of the battery is initially determined. Then, each disassembly node is further analyzed, and the disassembly feasibility of each node is comprehensively evaluated based on the connection method and disassembly capability of each node. Next, based on the initially determined disassembly sequence and the disassembly analysis results of each node, the disassembly feasibility of the battery is comprehensively evaluated. Finally, based on the disassembly feasibility of the battery, the disassembly sequence of the disassembly of the disassembled batteries is determined, and the non-disassembled batteries are crushed as a whole or reused in a tiered manner.

[0115] The specific process for dismantling and pre-processing retired power batteries is as follows:

[0116] First, the CAD model of the retired power battery is analyzed. Based on the three-dimensional spatial characteristics of the retired power battery in the CAD drawings and the disassembly experience of the retired power battery, a disassembly tree for the retired power battery is obtained. Then, based on the disassembly tree of the retired power battery, the disassembly sequence of the battery is preliminarily determined.

[0117]

[0118] Further analysis of each disassembly node:

[0119] (1) Analysis of the connection method of retired power battery nodes

[0120]

[0121] In the formula, c ii N represents the elements on the main diagonal of matrix C. All other elements in matrix C are assumed to be 0. j This indicates the number of nodes. When the connection method of the retired power battery node is bolt connection, spring connection, or mechanical pressing, the node is considered easy to disassemble. When the connection method is welding connection, the node is considered difficult to disassemble. When the key nodes related to the battery unit are difficult to disassemble, the disassembly and recycling of the retired power battery is not very meaningful. In this case, matrix C is assigned the value of zero matrix.

[0122] (2) Analyze the real-time status of retired power battery nodes.

[0123]

[0124] In the formula, d ii N represents the elements on the main diagonal of matrix C. Other elements in matrix D are assumed to be 0. j Indicates the number of nodes; when the node's shape... When the deformation exceeds the deformation threshold by 10%, the node is considered to have a large degree of rusting and deformation. When the deformation Δ of the key node related to the battery cell exceeds the deformation threshold, it is not very meaningful for the dismantling and recycling of the retired power battery. At this time, matrix D is assigned the value of zero matrix.

[0125] Taking into account the retired power battery node situation reflected by matrices C and D, when r(C) + r(D) ≥ 2N j When the value is -3 and r(C)≠0, r(D)≠0, the retired power battery is considered to be disassembled. Here, r(C) and r(D) are the ranks of matrices C and D, respectively.

[0126] Finally, based on the disassembly capability of the retired power battery, adjustments were made to the initially determined disassembly sequence. For example, if the connection method and node status of the battery chip's related nodes could not meet the disassembly conditions, the subsequent softening and disassembly operation could not be performed. In summary, the final disassembly sequence was obtained:

[0127]

[0128] Example 3

[0129] Disassembly Task Planning Module: In order to better complete the disassembly task, the disassembly task needs to be allocated according to the capabilities of the operator and the robot, giving full play to the advantages of both, thereby improving the disassembly efficiency.

[0130] like Figure 4 As shown, step 401 involves determining the disassembly sequence and the disassembly capabilities of the operator and robot; step 402 involves constructing a dynamic Bayesian network and converting it into a Hidden Markov Model (HMM) based on the characteristics of discrete Bayesian networks; step 403 involves improving the HMM based on the actual situation; and step 404 involves dynamically allocating disassembly tasks. The specific process includes:

[0131] An improved Hidden Markov Model is constructed. In the task allocation for dismantling retired power batteries, the operator working at time t is defined as the hidden sequence Q of the improved Hidden Markov Model, and the operation in progress at time t is defined as the observation sequence V of the improved Hidden Markov Model.

[0132]

[0133] V = {v1, v2, ... v} M}#(6)

[0134] Next, the key parameters of the improved Hidden Markov Model are determined, λ = (π, A, B). π represents the initial state probability vector, which is crucial in the task allocation for dismantling retired power batteries. i This represents the initial probability of selecting either a human operator or a robot to perform the disassembly task.

[0135] π=(π i ),π i =P(i1=q1), i=1,2,…,N o #(7)

[0136] A is from state q i Transition to state q j The state transition probability matrix, in the task allocation for dismantling retired power batteries, represents the state transition probability matrix at time t for operator a. ij At time t+1, B represents the probability of an operator (including itself); B is the probability from state q. j Observing a specific state v k The probability distribution matrix, in the allocation of tasks for dismantling retired power batteries, represents the probability distribution matrix when the operator is b. ij At that time, b ij The ongoing disassembly task represents the probability of a certain disassembly task.

[0137] Classical Hidden Markov Models typically use maximum likelihood estimation to determine model parameters. However, in Hidden Markov Models for power battery disassembly, maximum likelihood estimation is susceptible to overfitting. Therefore, this invention employs an improved Hidden Markov Model, combining expert knowledge or experience, and utilizes parameter estimation based on Bayes' theorem to determine the state transition probability matrix and probability distribution matrix.

[0138] (1) Define the prior probability density function

[0139] Based on expert knowledge or experience, define the prior probability density functions of the state transition probability matrix and the probability distribution matrix:

[0140] p(A|α d )#(8)

[0141] p(B|β d )#(9)

[0142] Where α d β d It is a probability distribution for the dismantling of retired power batteries derived from expert knowledge or experience.

[0143] (2) Calculate the posterior probability density function

[0144]

[0145] Where p(V|A,B) represents the probability of the observed sequence V given model parameters A and B, i.e., during the dismantling of retired power batteries, given time t, the operator a... ij At time t+1, let A be the state probability transition matrix of an operator (including itself) and when the operator is b.ij At that time, b ij When the ongoing dismantling task is the probability distribution matrix B of a certain dismantling task, the observation sequence V represents the probability of the dismantling operation in progress at time t.

[0146] Next, the posterior probability density function is normalized:

[0147]

[0148] Among them, A i B j Representing the possible values ​​of A and B, after normalization, we get This means that when given an observation sequence V during the dismantling of retired power batteries, A and B take different values.

[0149] (3) Determine the state transition probability matrix A and the probability distribution matrix B based on the posterior probability density function.

[0150]

[0151] Where, α t (i,j) represents the posterior probability that the user is in state i at time t and in state j at time t+1, where N is the posterior probability. o This indicates the number of elements in the hidden sequence, which is the number of operators involved in the dismantling of retired power batteries.

[0152]

[0153] Among them, v k β represents the k-th observation in the observation sequence. t (i) represents the posterior probability of being in state i at time t, N o M represents the number of elements in the hidden sequence, i.e., the number of operators in the dismantling process of retired power batteries, and M represents the number of elements in the observed sequence, i.e., the number of dismantling tasks in the dismantling process of retired power batteries.

[0154] Next, the task allocation is used as the hidden sequence Q, and the decomposition sequence output by the 301 decomposition preprocessing module is used as the observation sequence V. The optimal task allocation is then calculated using the Viterbi algorithm.

[0155] The Hidden Markov Model (HMM) is input with parameters λ = (π, A, B) and the observation sequence V. The Viterbi algorithm is then used to decode and analyze the model, ultimately yielding the optimal task allocation. As an improvement, constraint information between states is introduced during the Viterbi algorithm's recursion process to eliminate some unreasonable task allocations and simplify computation.

[0156] (1) Initialization

[0157] δ1(i)=πi b i (v1), i = 1, 2, ..., N o #(14)

[0158] ψ1(i)=0,i=1,2,…,N o #(15)

[0159] (2) Recursion (t = 2, 3, ..., T):

[0160] Define constraint matrix P

[0161]

[0162] Where, N m The number of decomposition operations is indicated by the constraint matrix P, which is then introduced into the recursive process.

[0163]

[0164]

[0165] (3) Termination:

[0166]

[0167]

[0168] Where, δ t (i) represents the maximum probability of all individual paths to state i at time t, i.e., the operator at time t in the dismantling process of the retired power battery is a. ij At that time, the maximum probability of all individual tasks being assigned a path; ψ t (i) represents the (t-1)th node of the path with the highest probability among all individual paths at time t, i.e., the operator in the dismantling process of the retired power battery at time t is a. ij At time t-1, the operator a of the path with the highest probability among all individual task assignment paths. i′j′ .

[0169] Disassembly simulation module: Based on the disassembly task allocation calculated by the dynamic algorithm, disassembly simulation is carried out on the Unity-based digital twin platform. Specific disassembly tasks are assigned to operators and robots in the virtual environment to simulate the actual disassembly process, eliminate potential safety hazards, and determine the feasibility of the solution.

[0170] Example 4

[0171] In this invention, combined with Figure 5Let's take EV battery disassembly as an example to demonstrate the feasibility of this invention through task allocation (500). In the diagram, 501 represents the disassembly sequence, 502 represents Bayesian network analysis, and 503 represents task allocation. The specific process is as follows:

[0172] The disassembly analysis first acquires the 3D features of the EV battery, i.e., the lithium iron phosphate battery. Data is transmitted to a digital twin platform via sensors to build a virtual environment for disassembling the lithium iron phosphate battery, generating a digital model. Next, a disassembly analysis algorithm is called to analyze the disassembly feasibility of the lithium iron phosphate battery and generate a disassembly sequence. After obtaining the disassembly sequence, a dynamic Bayesian network is constructed for analysis. This network is a discrete Bayesian network; to simplify calculations, a hidden Markov model is used to analyze the disassembly sequence.

[0173] First, a hidden Markov model is constructed, with the operator at time t as the hidden sequence Q and the operation in progress at time t as the observed sequence V.

[0174]

[0175] V = {v1, v2, ... v} M}#(twenty two)

[0176] Next, the key parameters of the Hidden Markov Model, λ = (π, A, B), are determined.

[0177] Based on the degree of automation in the disassembly sequence, an initial probability distribution π is obtained. For disassembling lithium iron phosphate batteries, precision tasks such as removing connecting wires and battery chips require manual operation and account for a small proportion. Tasks such as removing the battery pack casing and battery cells can be completed by robots and account for a larger proportion. Therefore, the initial probability distribution can be obtained as follows:

[0178] π = (0.2, 0.2, 0.3, 0.3) T #(twenty three)

[0179] Based on formulas (8)-(13), and taking into account the working positions of human operators and robots on the production line, the relationship between dismantling tasks, the dismantling capabilities of human operators and robots, and the dismantling tasks handled in daily work, the state transition matrix A and the probability distribution matrix B are obtained, taking into account the working positions of human operators and robots on the production line, the relationship between dismantling tasks, the dismantling capabilities of human operators and robots, and the dismantling tasks handled in daily work.

[0180]

[0181]

[0182] Next, the optimal task allocation is calculated using the Viterbi algorithm, with the input model parameters λ = (π, A, B) and the observation sequence.

[0183] V = {Remove casing A, disconnect connecting wire 2, remove battery unit A, remove battery chip A} #(26)

[0184] According to formulas (14)-(20), the Viterbi algorithm is used to find the optimal task allocation. Based on the local state recursion formulas (17) and (18), the maximum δ at time T is calculated. T (i) is the probability of the most likely hidden state occurring, i.e., the probability of optimal task allocation during EV battery disassembly; calculate the maximum ψ at time T. T (i) is the most likely hidden state at time T, i.e., the optimal task allocation at time T. Finally, ψ is used. T (i) Backtracking yields the most likely hidden sequence Q, which is the optimal task assignment:

[0185] Q = {Robot 1, Operator 1, Robot 1, Robot 2} #(27)

[0186] The above provides a detailed description of the human-machine collaborative dismantling method and application system for retired power batteries based on digital twin assistance provided by this invention. This invention innovatively utilizes a digital twin data platform to solve problems such as real-time monitoring of the dismantling process of retired power batteries and dynamic allocation of dismantling tasks, thereby improving dismantling efficiency.

[0187] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A human-machine collaborative dismantling method for retired power batteries based on digital twin assistance, characterized in that, The method includes the following steps: S101. Acquire 3D feature data of the physical model of the retired power battery and robot by image capture, and transmit the 3D feature data to the digital twin platform. S102. Construct a digital model based on 3D feature data, and achieve real-time monitoring of the physical model through real-time data fusion; S103. Based on the CAD model of the retired power battery, call the disassembly analysis algorithm to obtain the basic disassembly information of the battery, evaluate the disassembly feasibility of the battery, and determine the disassembly sequence. The method in step S103 includes: The digital twin platform will utilize disassembly analysis algorithms to analyze the CAD model of the retired power battery and obtain its disassembly tree. Based on this tree, a preliminary disassembly sequence will be determined. Further analysis of each disassembly node will then be conducted, and its disassembly capability will be comprehensively evaluated based on its connection method and disassembly capacity. Following this, the overall disassembly capability of the battery will be comprehensively evaluated based on the preliminary disassembly sequence and the analysis results of the disassembly capability of each node. Finally, based on the battery's disassembly capability, the disassembly sequence of removable batteries will be determined, and non-removable batteries will be either crushed as a whole or reused in a tiered manner. The method in step S103 specifically includes: (1) Analyze the CAD model of the retired power battery, and based on the three-dimensional spatial characteristics of the retired power battery in the CAD drawings and the disassembly experience of the retired power battery, obtain the disassembly tree of the retired power battery; then, based on the disassembly tree of the retired power battery, preliminarily determine the disassembly sequence of the battery: (2) Further analyze each dismantling node and the connection method of the retired power battery node: In the formula, N represents the elements on the main diagonal of matrix C. All other elements in matrix C are assumed to be 0. j This indicates the number of nodes; the methods for determining the difficulty of node decomposition include: When the connection method of the retired power battery node is bolt connection, spring connection, or mechanical pressing, the node is considered easy to disassemble; when the connection method is welding connection, the node is considered difficult to disassemble; when the key nodes related to the battery unit are difficult to disassemble, the disassembly and recycling of the retired power battery is not very meaningful, and matrix C is assigned the value of zero matrix. (3) Analyze the real-time status of retired power battery nodes: In the formula, N represents the elements on the main diagonal of matrix C. Other elements in matrix D are assumed to be 0. j Indicates the number of nodes; when the node's shape... When the deformation exceeds the deformation threshold by 10%, the node is considered to have a high degree of rusting and deformation; when the deformation of key nodes related to the battery cell exceeds the threshold... When the deformation threshold is exceeded, the dismantling and recycling of the retired power battery is not very meaningful. At this time, matrix D is assigned the value of zero matrix. (4) Taking into account the retired power battery node situation reflected by matrices C and D, when At that time, it was believed that the retired power battery could be disassembled; among them, Let C and D be the ranks of matrix C and matrix D, respectively. (5) Based on the disassembly of the retired power battery, the initially determined disassembly sequence is adjusted to obtain the final disassembly sequence; S104. Based on the disassembly sequence and the disassembly capabilities of the operator and the robot, construct a dynamic Bayesian network to dynamically allocate disassembly tasks and obtain the optimal task allocation. S105. Based on the optimal task allocation, conduct a breakdown simulation on the digital twin platform to determine the feasibility of the optimal task allocation. S106. Based on the disassembly simulation results, plan the robot's motion path and complete the human-machine collaborative disassembly.

2. The human-machine collaborative dismantling method for retired power batteries based on digital twin assistance according to claim 1, characterized in that, The method in step S101 includes: In the real world, 3D feature data of physical models of retired power batteries and robots are collected by 3D motion-sensing cameras and transmitted to a digital twin platform based on the Unity game engine. In a digital twin platform, virtual models and virtual scenes are constructed based on the 3D feature data of the physical model. Unity's scripting capabilities are used to write code to achieve data interaction, data analysis, and real-time data monitoring.

3. The human-machine collaborative dismantling method for retired power batteries based on digital twin assistance according to claim 1, characterized in that, The method in step S104 includes: Based on the disassembly sequence and the disassembly capabilities of the operator and the robot, a dynamic Bayesian network is constructed. According to the characteristics of discrete Bayesian networks, the dynamic Bayesian network is converted into a hidden Markov model. The hidden Markov model is then improved according to the actual situation, and the disassembly tasks are dynamically allocated. An improved Hidden Markov Model (HMM) is constructed. In the task allocation for dismantling retired power batteries, the operator working at time t is defined as the hidden sequence Q of the improved HMM, and the operation in progress at time t is defined as the observation sequence V of the improved HMM. The key parameters of the improved Hidden Markov Model are determined, λ=(π,A,B); π represents the initial state probability vector, which is crucial in the task allocation for dismantling retired power batteries. This represents the initial probability of selecting either a human operator or a robot to perform the disassembly task; A is from the state Transition to state The state transition probability matrix, in the task allocation for dismantling retired power batteries, represents the operator at time t. At time t+1, B represents the probability of a certain operator; B is derived from the state. Observing a specific state The probability distribution matrix, in the allocation of tasks for dismantling retired power batteries, represents the probability distribution matrix when the operator is... hour, The ongoing dismantling task represents the probability of a certain dismantling task. An improved Hidden Markov Model is employed, combined with expert knowledge or experience, and parameter estimation based on Bayes' theorem is used to determine the state transition probability matrix and probability distribution matrix: (1) Define the prior probability density function Based on expert knowledge or experience, define the prior probability density functions of the state transition probability matrix and the probability distribution matrix: in , It is a probability distribution for the dismantling of retired power batteries derived from expert knowledge or experience; (2) Calculate the posterior probability density function in This represents the probability of the observation sequence V given model parameters A and B, i.e., during the dismantling of retired power batteries, at a given time t, the operator... At time t+1, the state probability transition matrix A of a certain operator is... hour, When the ongoing dismantling task is the probability distribution matrix B of a certain dismantling task, the observation sequence V is the probability of the dismantling operation in progress at time t. Normalize the posterior probability density function: in, , Represents the possible values ​​of A and B, and after normalization, we get , which means the probability that A and B take different values ​​when given an observation sequence V during the dismantling of retired power batteries; (3) Determine the state transition probability matrix A and the probability distribution matrix B based on the posterior probability density function: in, N represents the posterior probability of being in state i at time t and in state j at time t+1. o This indicates the number of elements in the hidden sequence, i.e., the number of operators involved in the dismantling of retired power batteries; in, This represents the k-th observation in the observation sequence. N represents the posterior probability of being in state i at time t. o M represents the number of elements in the hidden sequence, i.e., the number of operators in the dismantling process of retired power batteries; M represents the number of elements in the observed sequence, i.e., the number of dismantling tasks in the dismantling process of retired power batteries. Next, the task allocation is used as the hidden sequence Q, and the output decomposition sequence is used as the observation sequence V. The Viterbi algorithm is then used to calculate the optimal task allocation.

4. The human-machine collaborative dismantling method for retired power batteries based on digital twin assistance according to claim 3, characterized in that, Specific methods for calculating the optimal task allocation using the Viterbi algorithm include: The Hidden Markov Model (HMM) is input with parameters λ=(π,A,B) and observation sequence V. The Viterbi algorithm is then used to decode and analyze the model, ultimately obtaining the optimal task allocation. As an improvement, constraint information between states is introduced during the Viterbi algorithm's recursion process to eliminate some unreasonable task allocations and simplify computation. This includes: (1) Initialization: (2) Recursively calculate t=2,3,…,T: Define constraint matrix P Where, N m The number of decomposition operations is indicated by the constraint matrix P, which is then introduced into the recursive process. (3) Termination: in, This represents the maximum probability of state i at time t, i.e., the operator at time t during the dismantling process of the retired power battery is... At that time, the maximum probability of all individual tasks being assigned a path; This represents the (t-1)th node of the path with the highest probability among all individual paths at time t, i.e., the operator in the dismantling process of the retired power battery at time t is... At time t-1, the operator of the path with the highest probability among all individual task assignment paths. .

5. The human-machine collaborative dismantling method for retired power batteries based on digital twin assistance according to claim 4, characterized in that, The method in step S105 includes: The maximum value at time T is calculated. It is the probability of the most likely hidden state occurring, i.e., the probability of optimal task allocation during battery disassembly; calculate the maximum probability at time T. It is the most likely latent state at time T, i.e., the optimal task allocation at time T; finally, it utilizes... Backtracking ultimately yields the most likely hidden sequence Q, which is the optimal task assignment: 。 6. A human-machine collaborative dismantling system for retired power batteries based on digital twin assistance, characterized in that, include: The physical model in the real environment is used to acquire 3D feature data of the physical model of the retired power battery and robot through image capture, and the 3D feature data is transmitted to the digital twin platform. The digital twin platform includes: Virtual models are used to interact with physical models in real time, enabling real-time monitoring of the physical models through real-time data fusion. The disassembly analysis algorithm is used to obtain basic disassembly information of a retired power battery based on its CAD model, assess its disassembly feasibility, and determine the disassembly sequence. Based on the disassembly sequence and the disassembly capabilities of the operator and robot, a dynamic Bayesian network is constructed to dynamically allocate disassembly tasks and obtain the optimal task allocation. Among these: The digital twin platform will utilize disassembly analysis algorithms to analyze the CAD model of the retired power battery, obtaining its disassembly tree. Based on this tree, a preliminary disassembly sequence will be determined. Further analysis of each disassembly node will then be conducted, comprehensively evaluating its disassembly feasibility based on its connection method and disassembly capability. Following this, the overall disassembly feasibility of the battery will be comprehensively evaluated based on the preliminarily determined disassembly sequence and the analysis results of each node's disassembly feasibility. Finally, based on the battery's disassembly feasibility, the disassembly sequence for removable batteries will be determined, and non-removable batteries will be either crushed as a whole or reused in a tiered manner. Specifically: (1) Analyze the CAD model of the retired power battery, and based on the three-dimensional spatial characteristics of the retired power battery in the CAD drawings and the disassembly experience of the retired power battery, obtain the disassembly tree of the retired power battery; then, based on the disassembly tree of the retired power battery, preliminarily determine the disassembly sequence of the battery: (2) Further analyze each dismantling node and the connection method of the retired power battery node: In the formula, N represents the elements on the main diagonal of matrix C. All other elements in matrix C are assumed to be 0. j This indicates the number of nodes; the methods for determining the difficulty of node decomposition include: When the connection method of the retired power battery node is bolt connection, spring connection, or mechanical pressing, the node is considered easy to disassemble; when the connection method is welding connection, the node is considered difficult to disassemble; when the key nodes related to the battery unit are difficult to disassemble, the disassembly and recycling of the retired power battery is not very meaningful, and matrix C is assigned the value of zero matrix. (3) Analyze the real-time status of retired power battery nodes: In the formula, N represents the elements on the main diagonal of matrix C. Other elements in matrix D are assumed to be 0. j Indicates the number of nodes; when the node's shape... When the deformation exceeds the deformation threshold by 10%, the node is considered to have a high degree of rusting and deformation; when the deformation of key nodes related to the battery cell exceeds the threshold... When the deformation threshold is exceeded, the dismantling and recycling of the retired power battery is not very meaningful. At this time, matrix D is assigned the value of zero matrix. (4) Taking into account the retired power battery node situation reflected by matrices C and D, when At that time, it was believed that the retired power battery could be disassembled; among them, Let C and D be the ranks of matrix C and matrix D, respectively. (5) Based on the disassembly of the retired power battery, the initially determined disassembly sequence is adjusted to obtain the final disassembly sequence; Disassembly simulation: Based on the optimal task allocation, disassembly simulation is conducted on the digital twin platform to determine the feasibility of the optimal task allocation; based on the disassembly simulation results, the robot's motion path is planned to complete the human-machine collaborative disassembly.

7. The human-machine collaborative dismantling system for retired power batteries based on digital twin assistance as described in claim 6, characterized in that, The digital twin platform includes a disassembly preprocessing module, which calls a disassembly analysis algorithm to analyze the CAD model of the retired power battery and obtain its disassembly tree. Based on this tree, the disassembly sequence is preliminarily determined. Then, each disassembly node is further analyzed, and its disassembly capability is comprehensively evaluated based on its connection method and disassembly capacity. Next, based on the preliminarily determined disassembly sequence and the analysis results of the disassembly capability of each node, the overall disassembly capability of the battery is comprehensively evaluated. Finally, based on the battery's disassembly capability, the disassembly sequence of removable batteries is determined, and non-removable batteries are either crushed as a whole or reused in a tiered manner.

8. The human-machine collaborative dismantling system for retired power batteries based on digital twin assistance according to claim 7, characterized in that, The digital twin platform includes a task planning module, which is used to construct a dynamic Bayesian network based on the disassembly sequence and the disassembly capabilities of the operator and the robot. Based on the characteristics of discrete Bayesian networks, the dynamic Bayesian network is converted into a hidden Markov model, and the hidden Markov model is improved according to the actual situation to dynamically allocate disassembly tasks. The digital twin platform includes a disassembly simulation module, which is used to perform disassembly simulation based on the disassembly task allocation calculated by a dynamic algorithm. The module assigns specific disassembly tasks to operators and robots in the virtual environment, simulates the actual disassembly process, eliminates potential safety hazards, and determines the feasibility of the optimal task allocation.

Images